Key takeaways
- Google Search Console lumps AI Mode and AI Overviews into the same "Web" search type, so you need workarounds (regex query filters, position-jump alerts) or a third-party tool to tell them apart
- AI Mode is conversational and real-time (Shopping Graph, Maps, live data), while AI Overviews is a static, one-shot summary built on pre-indexed content, meaning they often reward different pages for the same topic
- Query fan-out is the mechanism behind both features. A single prompt can trigger 5 to 20+ background searches, and Surfer's analysis found that 67.82% of AI Overview citations came from pages that weren't in the top 10 for the main query at all
- Google AI Overviews cites around 10 sources per answer, roughly double ChatGPT's average of 5, which matters when you're deciding how many "slots" realistically exist to compete for
- Single-snapshot rank checks are close to meaningless for AI features since responses vary run to run; you need repeated, scheduled tracking, not a one-off check
Why AI Mode and AI Overviews need separate attention in the first place
I'll be honest, when I first started tracking these two features I treated them as basically the same thing with different names. That was a mistake. They're built differently, they show up differently in your reporting, and they often cite entirely different pages for the exact same query.
Google's own documentation is pretty blunt about this: "AI Mode and AI Overviews may use different models and techniques, so the set of responses and links they show will vary." AI Overviews only appears when Google's systems decide it's additive to the classic results, and it frequently doesn't trigger at all. AI Mode, on the other hand, is the full conversational search experience, chat-like, multi-step, pulling in real-time data like stock prices, restaurant availability, and the Shopping Graph. It's currently English-only but live in 180+ countries, while AI Overviews covers 200+ countries and 40+ languages. One rolls out broader, one rolls out deeper.
Keyword.com put it well: "AI Mode changes how you search... AI Overviews change what you see." That distinction has practical consequences for anyone trying to monitor visibility, because the two features reward different content shapes, and they get buried together in the same reporting bucket.
The reporting problem: Search Console doesn't separate them cleanly
This is the part that trips most people up. Since June 2025, AI Mode data has technically been available in Search Console, Performance → Search Results → add a filter → Search Appearance → "AI Mode." Good, right? Except both AI Mode and AI Overviews get reported inside the standard "Web" search type, with no way to fully isolate AI Overviews from classic organic results in the default view. There's a Search Appearance tag for AI Mode, but Google "does not currently provide a separate AI Mode filter" for cleanly peeling it out of everything else.
A few other quirks worth knowing before you build any kind of dashboard on top of GSC data:
- When your page is cited inside an AI Overview, the impression inherits the AI Overview block's position, usually #1, but only registers once a user actually scrolls the citation into view
- AI Mode follow-up queries (the second, third message in a conversation) get attributed to the follow-up query text itself, not the original question that started the session
- If your content gets cited in an AI Overview but nobody clicks, Search Console shows you nothing. You could be the primary source for hundreds of queries and have zero visible evidence
A workaround a few practitioners use: filter the Query dimension in GSC with a regex for long, conversational queries (six-plus words: ^(?:\S+\s+){5,}\S+$, or ten-plus words for a stricter cut) since those are far more likely to be AI Mode fan-out queries than typical head-term searches. Also watch for keywords where your average position suddenly jumps to #1 with no corresponding content or backlink change, that's often a sign the page just got pulled into a citation.
GA4 has its own blind spot
GA4 doesn't fare much better. AI Overview clicks show up as plain old Google organic traffic, there's no referral parameter distinguishing them. GA4's built-in "AI Search" channel mostly catches ChatGPT and Perplexity referrals, not AIO traffic, because Google doesn't pass that signal.
A common fix is to build a custom channel group: Admin → Data Display → Channel Groups → create an "AI Search" channel with a condition matching Source against a regex like google.*ai|chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com, then move it above Organic Search and Referral in priority order. It's not perfect for isolating AI Overviews specifically, but it at least separates the broader AI-driven traffic bucket from everything else. Worth noting: referral traffic from Perplexity has been shown to convert at roughly 10.5%, about six times the typical 1.76% conversion rate of plain Google organic, so even imperfect segmentation here pays off.
Understanding query fan-out (because it drives both features)
If you want to actually influence AI Mode and AI Overviews rather than just watch them, you need to understand query fan-out. At Google I/O 2025, Head of Search Elizabeth Reid described how AI Mode "calls on our custom version of Gemini to break the question into different subtopics, and it issues a multitude of queries simultaneously on your behalf." Estimates put a single AI Mode question at generating anywhere from 5 to 20+ sub-queries depending on complexity, and Deep Search pushes that into the hundreds for a single report.
Here's the number that should change how you think about ranking for these features: Surfer SEO analyzed 173,902 URLs across 10,000 keywords and found that 67.82% of pages cited in AI Overviews were NOT in the top 10 organic results for the main query OR any of its sub-queries. Ranking #1 for the head term gets you cited only 19.6% of the time on its own. But ranking for the main query plus at least one sub-query accounts for 51.2% of citations. In other words, the fight isn't just for the obvious keyword, it's for the cloud of related sub-questions AI Mode and AI Overviews generate behind the scenes.
Promptwatch's own fan-out research on ChatGPT shows a parallel trend worth keeping in mind: average fanout queries per response dropped from about 2.15 in December 2025 to roughly 1.84 by March 2026, and query length shrank from ~117 characters to ~53 characters over the same period, meaning fan-out queries are getting shorter and more keyword-like across AI search generally. The methodology for tracking this (https://promptwatch.com/data/chatgpt-query-fanouts) is the same kind of lens worth applying to AI Mode's fan-out behavior, even though Google hasn't published exact per-query counts.
What content actually gets cited, and how that's shifting
If you're deciding what to optimize first, it helps to know what AI Overviews is favoring right now. Promptwatch's July 2026 data on AI Overviews citation types (https://promptwatch.com/data/ai-overviews-citation-types-over-time-july-2026) shows product pages overtaking listicles as the most-cited format from July 28 onward, ending the month at 17.9% versus 16.2% for listicles. That's a real shift: listicles averaged around 26% of citations back in Q1 2026, dropping to about 18% by July, while product pages roughly doubled from 9% to 16-18% over the same stretch. Video citations also climbed, from about 2.7% in January to nearly 6.3% by late July.
Source volume matters too. Promptwatch's cross-engine comparison (https://promptwatch.com/data/average-sources-per-response) found Google AI Overviews cites roughly 10 sources per answer, consistently, while ChatGPT cites around 5. Perplexity sits close to 10 as well, with very little variance. Microsoft Copilot is the outlier, swinging from under 2 to nearly 17 sources per response within weeks, which tells you Microsoft is still re-architecting its retrieval setup. If you're building out a tracking dashboard, this is useful context for setting expectations: AI Overviews realistically has twice as many "citation slots" open per response as ChatGPT does.
And on AI Mode specifically, Promptwatch's June 2026 citation-share data (https://promptwatch.com/data/google-ai-mode-citation-share-june-2026) found Google's own properties, including YouTube, now capture over 10% of AI Mode citations, with Google.com self-citations climbing from 4.24% in May to 7.31% in June, a 72% jump in a single month. Below the top three sources, every other independent domain drops under 1% share, which tells you the open-web opportunity in AI Mode lives almost entirely in a long tail of smaller sites, not a handful of dominant ones.
A combined monitoring workflow that actually works
Here's roughly how I'd structure this if I were setting it up from scratch:
1. Separate GSC signal cleanup. Set up the regex query filters for long-tail conversational queries, and build a saved view that tracks sudden position-to-#1 jumps. This gives you a rough, free proxy for AI feature activity without paying for anything.
2. GA4 channel segmentation. Build the custom "AI Search" channel group described above so AI-adjacent traffic at least gets bucketed separately from plain organic, even if it's imperfect.
3. A dedicated AI tracking tool for the parts GSC can't see. This is the real gap. GSC can't tell you whether a citation happened with zero clicks, can't separate AI Mode from AI Overviews cleanly, and can't show you competitor citations. A monitoring tool closes that gap.
4. Repeat runs, not snapshots. Rand Fishkin's SparkToro research found AI brand-recommendation lists are "so randomized that even if you ran the prompt hundreds of times, you'd likely never see the same list, in the same order, twice." A single rank check tells you almost nothing. You need scheduled, repeated tracking across dozens of runs to see a meaningful pattern, which is exactly what dedicated tools automate.
Choosing a tool: what actually separates AI Mode tracking from AI Overviews tracking
Not every "AI visibility" tool actually distinguishes AI Mode from AI Overviews, and some treat AI Mode as a paid add-on rather than a core feature. Here's how a few of the better-known options stack up:
| Tool | AI Mode included | AI Overviews included | Notable extras | Starting price |
|---|---|---|---|---|
| Otterly.ai | Paid add-on ($9-149/mo) | Yes, core | API/MCP access on Standard+ | $29/mo (Lite) |
| Semrush AI Visibility Toolkit | Yes, as platform filter | Yes | Topic/source gap reports | $99/mo |
| Ahrefs Brand Radar | Yes | Yes, per-platform index | 476M+ prompts tracked monthly | $199/mo per platform |
| SE Ranking Visible | Yes | Yes | AI Overview visibility tool on Core plan | $129/mo |
| Promptwatch | Yes, with crawler logs and citation analytics | Yes | Content Agents, Unified Actions, Reddit/YouTube citation tracking | $95/mo |
A quick note on methodology: Otterly explicitly runs "first-person" platform scans rather than raw API calls, which matters because personalization, like location, can skew what a manual search shows versus what a neutral scan captures. Whatever tool you pick, ask how it's actually pulling the data.
If you're also trying to close the loop between "here's where we're invisible" and "here's what to actually publish to fix it," Promptwatch is worth a look. It tracks AI Mode and AI Overviews alongside ChatGPT, Gemini, Perplexity, and the rest from one dashboard, but it also layers in AI crawler logs showing exactly when GoogleOther or Google-Agent crawls your pages, citation trend data broken into 22 content types, and a content gap analysis that turns those gaps into briefs or automatically published articles. That matters here specifically because, as the fan-out data above shows, winning AI Overview and AI Mode citations is about covering sub-query clusters, not just the head term, and that's a content-planning problem as much as a tracking problem.
Other tools worth evaluating depending on your setup: Otterly.AI if you want an affordable multi-engine starting point, Semrush if you already live in its SEO suite, and SE Ranking if you want AI Overview tracking bundled with traditional rank tracking at a lower price point.
Validating that a tool is actually tracking AI Mode (not just AI Overviews with a different label)
A surprising number of "AI Mode tracker" claims in vendor marketing are softer than they sound. Omnia's evaluation checklist for this is genuinely useful: confirm the tool returns different citations for different countries (proof that geo-localization is actually active, not hardcoded), check whether it captures citations at the URL level or only aggregates at the domain level, ask to see raw captured answer text with timestamps rather than just summary metrics, and verify it shows competitor citations, not just your own brand's.
Geo-level tracking matters more than people assume here. Brands expanding internationally often track only US AI Mode results while competitors are quietly winning visibility in UK or German markets that nobody's watching.
A realistic view of what this gets you
The honest takeaway is that no single free method gives you a clean, separated view of AI Mode and AI Overviews performance. GSC and GA4 get you partway there with some regex workarounds and custom channel groups, but the real gaps, zero-click citations, competitor visibility, sub-query coverage, only close with a dedicated tracking layer run on a schedule, not a one-off check. If you're exploring the wider category of GEO and AI visibility platforms beyond what's covered here, the directory at bestgeosoftware.com is a reasonable place to compare options side by side.